pyAdwin

ADWIN is an adaptive sliding window algorithm for detecting change and keeping updated statistics from a data stream, and use it as a black-box in place or counters in learning and mining algorithms initially not designed for drifting data. Reference: The Adwin Software

Implementation based on C ++ implementations in this repository Adwin C++

Prerequisites

• Python > 2.7

Install

• Clone this repository in your project folder

git clone https://github.com/rsdevigo/pyAdwin.git

• Access the pyAdwin folder and execute the setup.py

cd pyAdwin && sudo python setup.py install

Example

from pyadwin import Adwin

#Delta's standard of 0.01 , but Adwin builder can receive a floating point as the delta parameter.
adwin = Adwin()

data_stream = [1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.7]

for data in data_stream:
  if (adwin.update(data))
    print "Change has been detected in data: "+str(data)
    print adwin.getEstimation() # Prints the next value of the estimated form of data_stream